Epid 600 Class 2 Cause

EPID 600; Class 2 What is a cause? Causal inference and interpretation of epidemiologic evidence University of Michigan School of Public Health

Drug Abuse: A workshop on behavioral and economic research October 18-20, 2004 1

What is epidemiology? The study of the distribution and determinants of healthrelated states or events in specified populations, and the application of this study to the control of health problems

2

What is epidemiology? The study of the distribution and determinants of healthrelated states or events in specified populations, and the application of this study to control of health problems Therefore, epidemiology is fundamentally about the search for causes so that we may do something about them

3

Historical developments in the understanding of disease etiology The wrath of God “for now, I will stretch out mine hand, that I may smite thee and thy people with pestilence” God, from Exodus (9:14)

4

Historical developments in the understanding of disease etiology Rational thinking about disease causation started in 400 BCE with Hippocrates in the Epidemics Related symptoms of different illnesses to seasons and geography Focused on illnesses and sick persons as unique events

5

Historical developments in the understanding of disease etiology 1546: Fracastoro in “On contagion, contagious diseases and their treatment” Seminaria act on the humors of the body to create disease Three modes of transmission: person to person, fomites, airborne

6

Historical developments in the understanding of disease etiology 1600s Thomas Sydenham, “the English Hippocrates” “All diseases then ought to be reduced to certain and determinate kinds, with the same exactness as we see it done by botanic writers in their treatises of plants” Viewed diseases as distinct entities and began to hypothesize about causes William Petty and John Graunt First to use numerical data to describe patterns of mortality Proposed the establishment of a central government agency to collect data on vital information (Petty) Published Observations on the bills of mortality (Graunt) Analyzed records on causes of death from each parish 7

Historical developments in the understanding of disease etiology 1700s Giovanni Morgagni (“clinicopathologic correlation”) Associated certain signs and symptoms with specific pathologic changes in tissues and organs Spurred search for specific as opposed to general causes of diseases

8

Historical developments in the understanding of disease etiology 1800s 1831 Civil registration of vital status established in England 1838 England’s General Register Office established and headed by William Farr recorded all births and deaths; Farr developed disease classification system Zymotic (epidemic, endemic, contagious) Constitutional (gout, dropsy, cancer) Local (diseases of 8 organ systems) Developmental (diseases of childhood, old age, women, nutrition) Violent (accidents, battle deaths, homicides, suicides, executions, etc) 9

Historical developments in the understanding of disease etiology “Diseases which are communicated from person to person are caused by some material which passes from the sick to the healthy.” John Snow (1813-1858)

10

Historical developments in the understanding of disease etiology 1876: Koch reproducibly transmits anthrax to mice using the blood of infected cows Same rod-like material recovered from cows and

mice

Infection transmittable from mouse to mouse Koch’s postulates Proof that a particular microorganism is the cause of a particular infectious disease

11

Henle-Koch postulates 1. 

The agent must be present in every case of the disease

2. 

The agent must be isolated from the host and grown in vitro

3. 

The disease must be reproduced when a pure culture of the agent is inoculated into a healthy susceptible host

4. 

The same agent must be recovered once again from the experimentally infected host

12

Nelson, Williams, Graham. Infectious Disease Epidemiology Theory and Practice. Aspen Publishers, 2001

13

Three questions in causal inference 1. 

Methodological question? How do we look for a cause?

3. 

Ontological question What is a cause?

5. 

Ethical question? How do we decide if there is enough evidence to act on a cause?

14

1. The methodological question

15

Association vs. Causation Association is an identifiable relation between an exposure and a disease EXAMPLES Incidence rate of lung cancer is higher among smokers than among non-smokers Postmenopausal women on hormone replacement therapy (HRT) have lower rates of cardiovascular mortality than postmenopausal women who are not on HRT

16

Association vs causation Therefore if association is present we have to determine if exposure is truly a cause of disease EXAMPLES Does smoking cause lung cancer? Does HRT cause a reduction in the risk of death from cardiovascular diseases?

17

So how do we determine if something is causal?

18

When is an association causal?

Theory Hypothesis Strategy to test the hypothesis Design, conduct, and analysis of study Interpretation of results

19

When is an association causal?

Theory

Smoking is a carcinogen

Hypothesis Strategy to test the hypothesis Design, conduct, and analysis of study Interpretation of results

20

When is an association causal?

Theory

Smoking is a carcinogen

Hypothesis

Smoking causes lung cancer

Strategy to test the hypothesis Design, conduct, and analysis of study Interpretation of results

21

When is an association causal?

Theory

Smoking is a carcinogen

Hypothesis

Smoking causes lung cancer

Strategy to test the hypothesis

Prospective cohort study

Design, conduct, and analysis of study Interpretation of results

22

When is an association causal?

Theory

Smoking is a carcinogen

Hypothesis

Smoking causes lung cancer

Strategy to test the hypothesis

Prospective cohort study

Design, conduct, and analysis of study

Recruit 10,000 doctors, follow for 10 years

Interpretation of results

23

When is an association causal?

Theory

Smoking is a carcinogen

Hypothesis

Smoking causes lung cancer

Strategy to test the hypothesis

Prospective cohort study

Design, conduct, and analysis of study

Recruit 10,000 doctors, follow for 10 years

Interpretation of results

High RR of lung cancer in smokers

24

Induction vs. deduction Induction Specific observation

General premise

Deduction General premise

Specific observation

25

Confirmation vs. falsification Confirmation If A, then B, C, D B, C, D, therefore A

Falsification If A, then B, C, D NOT B, C, D, therefore NOT A

26

Observation A physician at a local hospital notices that she has seen two elderly patients presenting to the hospital in the span of a week with encephalitis (brain inflammation). She selects other patients to serve as “controls” and carries out a casecontrol study

27

Experiment Convinced by laboratory data and by observational data in humans, a group of scientists launch a trial of estrogen replacement therapy among post-menopausal women to explore if women taking hormone replacement therapy have better control of menopausal symptoms and fewer fractures

28

Advantages of observational vs. experimental studies OBSERVATION Cheaper

EXPERIMENT Variables of interest more readily controlled by investigator

Fewer ethical quandaries Faster to organize and conduct

Other extraneous variables more readily controlled by investigator

Can test multiple hypotheses and associations

29

Advantages of observational vs. experimental studies OBSERVATION Cheaper Fewer ethical quandaries Faster to organize and conduct

MAYBE

Can test multiple hypotheses and associations

30

Advantages of observational vs. experimental studies EXPERIMENT Variables of interest more readily controlled by investigator MAYBE

Other extraneous variables more readily controlled by investigator

31

A few well known causes of disease Smoking High cholesterol M. tuberculosis S. viridans Head injury Poverty [?]

32

A few well known causes of disease Smoking

Lung Cancer

High cholesterol

Cardiovascular Disease

M. tuberculosis

Tuberculosis

S. viridans

Endocarditis

Head injury

Subarachnoid hemorrhage

? Poverty

All-cause mortality

33

2. The ontological question

34

Criteria for Causal Inference (Bradford Hill 1965) Temporality Strength of association Biological plausibility Dose-response Replication of findings Consideration of alternate explanations Cessation of exposure Coherence with established facts Specificity of association

35

Temporality Exposure must precede disease In diseases with long latency periods, exposures must precede latency period In chronic diseases, often need long-term exposure for disease induction

36

Strength of association Strong associations are less likely to be caused by chance or bias A strong association means a very high or very low relative risk

37

Strength of association Strong associations are less likely to be caused by chance or bias A strong association means a very high or very low relative risk

CAVEAT Environmental associations with very low relative risks

38

Biologic plausibility The proposed mechanism should be biologically (etiologically) plausible Reference to a “coherent” body of knowledge

39

Biologic plausibility The proposed mechanism should be biologically (etiologically) plausible Reference to a “coherent” body of knowledge

CAVEAT High oxygen concentration causing neonatal retrolental fibroplasia

40

Dose-response relationship Changes in exposure are related to trend in risk of disease Strong evidence for causal relation suggesting biologic relation

41

Smith et al. 1994

30

27

25

22

20

17

15

,0

,5

,0

,5

,0

,5

,0

,5

,0

-3

-2

-2

-2

-2

-1

-1

-1

-1

9

9

9

9

9

9

9

9

9

49

49

99

49

99

49

99

49

99

2,

2,

9,

7,

4,

2,

9,

7,

4,

9

9

99

00

49

9,

,5

2,

0-

>3

00

00

00

00

00

00

00

00

00

50

Age-adjusted Mortality Rate (per 10,000 person-years)

12

10

7,

<7

Relation between income and mortality 90 80 70 60 50 40 30 20 10 0

$US 1980 42

Dose-response relationship Changes in exposure are related to trend in risk of disease Strong evidence for causal relation suggesting biologic relation

CAVEAT Thresholds, i.e., no disease past a certain level of exposure 43

Replication of findings Relations that are demonstrated in multiple studies are more likely to be causal Consistent results found in different populations, in different times, with different study designs

44

Replication of findings Relations that are demonstrated in multiple studies are more likely to be causal Consistent results found in different populations, in different times, with different study designs

CAVEAT Heterogeneity of effect in different countries

45

Consideration of alternate explanations Extent to which investigator has ruled out other possible explanations Methodologically sound studies with no potential residual confounding

46

Consideration of alternate explanations Extent to which investigator has ruled out other possible explanations Methodologically sound studies with no potential residual confounding

CAVEAT Alternate explanations limited by understanding of biology and sophistication of analysis 47

Cessation of exposure Risk of disease expected to decline when exposure to a cause is reduced or eliminated

48

Cessation of exposure Risk of disease expected to decline when exposure to a cause is reduced or eliminated

CAVEAT Pathogenic process already started; removal of cause does not reduce disease risk

49

Coherence with established “facts” If a relation is causal, would expect observed findings to be consistent with other data Hypothesized causal relations need to be consistent with epidemiologic and biologic knowledge

50

Coherence with established “facts” If a relation is causal, would expect observed findings to be consistent with other data Hypothesized causal relations need to be consistent with epidemiologic and biologic knowledge

CAVEAT Data may not be available yet to directly support proposed mechanism Science must be prepared to reinterpret existing understanding of disease process in the face of new evidence

51

Specificity of the association Specific exposure associated with only one disease Arises from old Henle-Koch postulates for causation

52

Specificity of the association Specific exposure associated with only one disease Arises from old Henle-Koch postulates for causation

CAVEAT Many exposures are linked to multiple diseases

53

Overall caveats to “criteria” “None of my ... [criteria] can bring undisputable evidence for or against the cause-and-effect hypothesis and none can be required as a sine qua non.”

54

Therefore, causal inference…  

Causal inference is not a simple (or quick) process

 

No single study is sufficient in establishing causal inference

 

Requires critical judgment and interpretation

 

Can one “prove” causality?

55

What is a cause? (Rothman) A cause of a disease is an event, condition, or characteristic that preceded the disease event and without which the disease event would not have occurred at all or would not have occurred until some later time.

56

Sufficient and component causes

U A

T B

X

Sufficient Cause 1

B

Sufficient Cause 2

A sufficient cause is a set of minimal conditions or events that inevitably produce disease 57

Sufficient and component causes

Component causes U A

T B

X

Sufficient Cause 1

B

Sufficient Cause 2

A sufficient cause is a set of minimal conditions or events that inevitably produce disease 58

Sufficient and component causes

A component cause is any one of a set of conditions which are necessary for the completion of a sufficient cause

Component causes U A

T B

X

Sufficient Cause 1

B

Sufficient Cause 2

A sufficient cause is a set of minimal conditions or events that inevitably produce disease 59

Sufficient and component causes

A necessary cause is a component cause that is a member of every sufficient cause

U A

T B

Sufficient Cause 1

X

B

Sufficient Cause 2 60

For example: Tuberculosis

Necessary but not sufficient

M. tuberculosis Poor nutrition

Immuno- M. tuberculosis suppression

Sufficient Cause 1 Neither necessary nor sufficient

Sufficient Cause 2 61

“Causing” a myocardial infarction

Potato chips Y W No exercise

62

“Causing” a myocardial infarction

Potato chips Y W No exercise

Obesity A

63

“Causing” a myocardial infarction

Potato chips Y W No exercise

Obesity A NO EFFECT

64

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

A C Genes

65

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C Genes

T

66

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C Genes

T NO EFFECT 67

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress

68

“Causing” a myocardial infarction Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress 69

Limitations of sufficient cause model  

Omits discussion of origins of causes, focuses on proximal causes

 

Specific components but not linkages among them

 

Does not consider factors that control distribution of risk factors

 

Ignores dynamic non-linear relations

70

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress71

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress

72

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress

73

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress

74

“Causing” a myocardial infarction

Potato chips Y W

Obesity

No exercise

High cholesterol

A C T

Genes Smoking

X

B Stress

75

Tuberculosis infection Tuberculosis is among the top ten causes of death in the world Tuberculosis is caused by infection with M. tuberculosis But knowing who is infected with M. tuberculosis does not necessarily inform us about the distribution of those with TB disease in populations…

76 http://www.who.int/mediacentre/factsheets/fs104/en/index.html

Tuberculosis infection About 2 billion people are infected with M. tuberculosis worldwide However, only 5-10% of those infected actually develop the disease So can we say that M. tuberculosis is the cause of TB?

http://www.who.int/mediacentre/factsheets/fs104/en/index.html

77

Vitamin D deficiency and TB Attack of macrophages is a critical step in the development of TB Vitamin D modulates monocyte-macrophage activity in the body Perhaps deficiencies in serum vitamin D levels cause TB? A meta-analysis conducted to evaluate the evidence

Nnoaham and Clarke. Int J Epidemiol. Low serum vitamin D levels and tuberculosis: a systematic review and meta-analysis. 2008; 37: 113-119

78

Vitamin D deficiency and TB

The underlying causal scenario of interest

Vitamin D

Risk of developing tuberculosis

Nnoaham and Clarke. Int J Epidemiol. Low serum vitamin D levels and tuberculosis: a systematic review and meta-analysis. 2008; 37: 113-119

79

Vitamin D deficiency and TB TB case mean vitamin D level (nmol/L)

Healthy control mean vitamin D level (nmol/L)

Study 1

16.0

27.25

Study 2

65.75

69.5

Study 3

39.75

65.5

Study 4

69.5

95.5

Study 5

46.5

52.25

Study 6

26.75

48.5

Is this enough evidence to call Vitamin D a cause of active TB? 80 Nnoaham and Clarke. Int J Epidemiol. Low serum vitamin D levels and tuberculosis: a systematic review and meta-analysis. 2008; 37: 113-119

Vitamin D deficiency and TB Do we have evidence to believe that the association is not in the reverse direction? (Temporal order) Vitamin D

Risk of developing tuberculosis

Do we have evidence to conclude that these cases would not have occurred but for vitamin D deficiency?

81

Counterfactual thinking Observed

Sick

Counterfactual (parallel universe)

Healthy

82

Counterfactual thinking Observed

Sick

Counterfactual (parallel universe)

Sick

83

The counterfactual universe

84

The counterfactual universe

D

No D

85

The counterfactual universe

D

No D

86

The counterfactual universe

D

No D

D

No D

87

The real universe

88

The real universe

D

No D

89

The real universe

E

No E

D

No D

90

The real universe

E

No E

D

No D

So the question, how are these two parts different? Are they “exchangeable”? Epidemiology is centrally concerned with ensuring exchangeability or comparability of these two parts of the population

91

3. Ethics and the public health balance  

When is there enough evidence to say something is a “cause”?

 

When should we decide that something is a cause and act on it?

 

Does “first do no harm” always apply at the population level?

 

Are there different guidelines for solutions where we have to DO something vs. solutions where we try to remove something?

92

Coda "All scientific work is incomplete - whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the action it appears to demand at a given time.“ Sir Austin Bradford Hill

93